Pinokio
FeedApps
Author@akos2
Download Pinokio
Sign inCreate accountJoin
Sign inCreate accountJoin

@akos2

Joined September 2026

5 Following0 Followers
PostsReplies
More
ActivityCheckpointsFollowed appsGitHub creations
Replies · 3
@rdx_aiHi, i havent seen any mlx implementation of ostris toolkit.. tbh, Lora training is very intensive and time consuming. I think Drawthings app has the fastest lora training currently for mac, but it doesnt have all the models support like in Ostris toolkit. BTW, i just checked, the mflux cli does also support lora train
@akos2·2d
in SeedVR2-7B MLX
It's not as intensive as you think. And it's also not that time consuming, because you can, you just let it run overnight, or you use Chivitai. I've done hundreds of LoRAs so far. I use Draw Things LoRAs, which are not the best. Chivitai was the best so far, and that's Koya. And then I've seen results with the toolkit thing that are way beyond what I have seen or experienced with Koya. But I've not done any LoRAs yet with toolkit. And it takes a huge amount of time. So if it would help you, I don't know what hardware you have. I just got new hardware. You're welcome to play with screen sharing, you know, for a few hours or so on my computer, and use mine if it helps you, you know, go further. But this would be amazing if you could consider this, because your app obviously is just light speed fast, and LoRA training could shorten when using MLX massively, and you would definitely spearhead the bleeding edge here of this tech. And I wouldn't be surprised if the guy behind the toolkit thing, this guy called Ostris, would give you a helping hand, because he's 100% PC, and he has no knowledge about Mac. And I'm probably the only one that ever asked him to also do a Mac version, and he reluctantly did one. It's on GitHub, and I think that's where Morpheus and Anchorite then branched or forked off and did the Pinocchio version. But I'm happy to test. I'm happy to do stuff here, because I kind of tinker with it all the time. And if your MFlux CLI supports LoRA training, then the question is it Koya, is it Toolkit, or is it the One Trainer? There are only three of these around, from what I understand. And then Draw Things uses PEFT, P-E-F-T. So the landscape is very narrow. That would be amazing if you could have a look at it and make a Pinocchio app, man. You'd become a hero in that kind of section of the community the second you would do this.
@rdx_aiUpdate v3 : Changelogs : 1. input image size for the uploaded image shown in the UI. 2. zoom,pan for the result tab . 3. model download progress bar in UI, background downloads support. 4. a paste from clipboard button( along with keyboard shortcut) to add images 5. option for random seed by default.
@akos2·3d
in SeedVR2-7B MLX
So cool, will try right away..... by the way , since you are the top expert with mps and mlx, would you know if there is an Ostris Ai Tollkit fork or branch yet out that uses mlx the way you did here in your mlx app? i see pinokio did 2 installers for ai toolkit, and the first install is failing and stopping during training.... and am trying to see if Morepheous an Anchorite can figure out why... but your speed here using MLX in yyour app is absurdly insane, and could / is desperately usefull in the lora training app named ai toolkit. this app must works for hours usually to train loras, and i can only phantom how much faster lora training could be, if ai toolkit were to use propper MLX use... just wanted to put it out there and mention this to you....
@rdx_ailooks like one of the libraries not getting installed. are there any more logs? could u send me the app logs( 'get help' also copies full logs). Also, did the models get downloaded?
@akos2·7d
in SeedVR2-7B MLX
tried again toda still not working, blodgy soft speckled image result, see sscreesnhot. i looked for what you suggested, the get help section, not sure if thsi is what you caan use or need: Get Help Log REPORT PREVIEW Markdown draft # Issue Report App: SeedVR2-7B MLX (SeedVR2-MLX.pinokio.git) Repo: github.com/Rdx-ai-art/SeedVR2-M… Generated: 2026-10-15T09:14:55.457Z Pinokio: 8.2.0 Platform: darwin arm64 Node: v22.21.1 ## Summary ## System ```json { "pinokio": { "version": "8.2.0", "node": "v22.21.1", "platform": "darwin", "arch": "arm64" }, "hardware": { "gpu": "apple", "gpu_model": "apple m5 max", "ram_gb": 128, "vram_gb": 0, "gpu_driver": null, "gpu_target": null }, "os": { "platform": "darwin", "distro": "macOS", "release": "27.0.1", "codename": "Golden Gate", "kernel": "27.0.0", "arch": "arm64", "build": "26A434", "uefi": true }, "system": { "manufacturer": "Apple Inc.", "model": "Mac17,14", "version": "Unknown", "virtual": false }, "cpu": { "manufacturer": "Apple", "brand": "M5 Max", "vendor": "Apple", "family": "-143893734", "stepping": "5", "speed": 2.4, "speedMin": 2.4, "speedMax": 2.4, "cores": 18, "physicalCores": 18, "processors": 1, "performanceCores": 6, "efficiencyCores": 12, "virtualization": true, "cache": { "l1d": 65536, "l1i": 131072, "l2": 8388608 } }, "memory": { "total": 137438953472, "free": 7480803328, "used": 129958150144, "active": 91797356544, "available": 45641596928, "buffers": 0, "cached": 0, "slab": 0, "buffcache": 38160793600, "swaptotal": 1073741824, "swapused": 62914560, "swapfree": 1010827264 }, "gpus": [ { "model": "apple m5 max" } ], "graphics": { "controllers": [ { "vendor": "Apple", "model": "Apple M5 Max", "bus": "Built-In", "vramDynamic": true, "cores": "40" } ], "displays": [ { "model": "S32B80P", "main": true, "builtin": false, "currentResX": 3008, "currentResY": 1692, "resolutionX": 6016, "resolutionY": 3384, "currentRefreshRate": 60 } ] }, "diagnostics": { "platform": "darwin", "arch": "arm64", "running": { "/Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/start.js": true }, "home": "/Users/akos/pinokio", "vars": { "pip": { "install": { "torch": "pip3 install torch torchvision torchaudio" } } }, "memory": { "local": { "/Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/start.js": { "url": "127.0.0.1:42003" } }, "global": {}, "rpc": { "/Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/start.js": { "method": "local.set", "params": { "url": "{{input.event[1]}}" } } }, "input": { "/Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/start.js": { "id": "18a874a6-e258-4c39-98b2-d91eb1b8ab76", "response": "\r\nThe default interactive shell is now zsh.\r\nTo update your account to use zsh, please run `chsh -s /bin/zsh`.\r\nFor more details, please visit support.apple.com/kb/HT208050.\…<<PINOKIO_SHELL>>eval \"$(conda shell.bash hook)\" ; conda deactivate ; conda deactivate ; conda deactivate ; conda activate base ; source /Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/app/env/bin/activate /Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/app/env && python app.py --port 42003\r\nMetal memory limit: 24 GB\r\n/Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/app/app.py:189: DeprecationWarning: \r\n on_event is deprecated, use lifespan event handlers instead.\r\n\r\n Read more about it in the\r\n [FastAPI docs for Lifespan Events](fastapi.tiangolo.com/advanced/e…).\r\n \r\n @app.on_event(\"startup\")\r\n 127.0.0.1:42003", "stdout": "\r\nThe default interactive shell is now zsh.\r\nTo update your account to use zsh, please run `chsh -s /bin/zsh`.\r\nFor more details, please visit support.apple.com/kb/HT208050.\…<<PINOKIO_SHELL>>eval \"$(conda shell.bash hook)\" ; conda deactivate ; conda deactivate ; conda deactivate ; conda activate base ; source /Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/app/env/bin/activate /Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/app/env && python app.py --port 42003\r\nMetal memory limit: 24 GB\r\n/Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/app/app.py:189: DeprecationWarning: \r\n on_event is deprecated, use lifespan event handlers instead.\r\n\r\n Read more about it in the\r\n [FastAPI docs for Lifespan Events](fastapi.tiangolo.com/advanced/e…).\r\n \r\n @app.on_event(\"startup\")\r\n 127.0.0.1:42003", "event": [ "127.0.0.1:42003", "127.0.0.1:42003" ] } }, "args": { "/Users/akos/pinokio/api/SeedVR2-MLX.pinokio.git/start.js": {} } }, "procs": {}, "gpu": "apple", "gpus": [ { "name": "apple", "model": "apple m5 max", "driver": null } ], "gpu_model": "apple m5 max", "gpu_driver": null, "gpu_target": null, "ram": 128, "vram": 0, "version": "8.2.0", "system": { "manufacturer": "Apple Inc.", "model": "Mac17,14", "version": "Unknown", "serial": "J7F903QDY2", "uuid": "6a26c3f8-9392-552d-9bd2-d9bbcc5dabcd", "sku": "J775cAP", "virtual": false, "type": "Other" }, "osInfo": { "platform": "darwin", "distro": "macOS", "release": "27.0.1", "codename": "Golden Gate", "kernel": "27.0.0", "arch": "arm64", "hostname": "Mac-Studio.local", "fqdn": "Mac-Studio.local", "codepage": "UTF-8", "logofile": "apple", "serial": "25FECA3D-AD9D-3958-B633-A14EC36D3BE7", "build": "26A434", "servicepack": "", "uefi": true }, "mem": { "total": 137438953472, "free": 7480803328, "used": 129958150144, "active": 91797356544, "available": 45641596928, "buffers": 0, "cached": 0, "slab": 0, "buffcache": 38160793600, "reclaimable": 49759469568, "swaptotal": 1073741824, "swapused": 62914560, "swapfree": 1010827264, "writeback": null, "dirty": null }, "shell": "/bin/bash", "cpu": { "manufacturer": "Apple", "brand": "M5 Max", "vendor": "Apple", "family": "-143893734", "model": "", "stepping": "5", "revision": "", "voltage": "", "speed": 2.4, "speedMin": 2.4, "speedMax": 2.4, "governor": "", "cores": 18, "physicalCores": 18, "performanceCores": 6, "efficiencyCores": 12, "processors": 1, "socket": "SOC", "flags": "", "virtualization": true, "cache": { "l1d": 65536, "l1i": 131072, "l2": 8388608, "l3": null } }, "audio": [ { "id": 0, "name": "S32B80P", "manufacturer": "SAM", "revision": null, "driver": null, "default": false, "channel": "Display-Port", "type": "Speaker", "in": false, "out": true, "status": "online" }, { "id": 1, "name": "Samsung SlimFit Cam", "manufacturer": "Alpha Imaging Tech. Corp.", "revision": null, "driver": null, "default": true, "channel": "USB", "type": "Microphone", "in": true, "out": false, "status": "online" }, { "id": 2, "name": "Mac Studio Speakers", "manufacturer": "Apple Inc.", "revision": null, "driver": null, "default": true, "channel": "Built-In", "type": "Speaker",
Show more
github.comGitHub - Rdx-ai-art/SeedVR2-MLX.pinokio: A MAC ONLY-MLX optimized Pinokio app for SeedVR2 upscale model using native MLX inferenceA MAC ONLY-MLX optimized Pinokio app for SeedVR2 upscale model using native MLX inference - Rdx-ai-art/SeedVR2-MLX.pinokio
You’re all caught up.
Pinokio
PrivacyTerms
FeedApps